An individualized optimization method and system for acoustic stimulation neuroregulation based on historical electroencephalogram and efficacy feedback

Abstract

The application discloses a sound stimulation neural regulation individual optimization method and system based on historical electroencephalogram and efficacy feedback. The method is based on historical electroencephalogram data, sound stimulation parameters, sound stimulation signal types, state variables, subjective feedback, efficacy indicators and electroencephalogram objective efficacy indicators obtained in the process of multiple sound stimulation interventions, establishes a parameter-signal-state-efficacy mapping relationship, constructs an individual response model, and generates individualized parameter recommendation range, sound stimulation signal type recommendation, parameter priority suggestion, parameter disabled range, parameter search space contraction result, efficacy trend prediction result or efficacy response probability. The model output needs to be confirmed by the doctor end, constrained by the prescription template and / or the safety state machine, and then used for subsequent parameter initialization, sound stimulation signal type selection, parameter update suggestion generation or parameter search space limitation. The application improves the individual adaptability, safety, interpretability and traceability of long-term sound stimulation neural regulation scheme.

Description

Technical Field

[0001] This invention relates to the field of medical device systems technology, and in particular to a method and system for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback. Background Technology

[0002] Acoustic stimulation neuromodulation is a non-invasive technique that inputs acoustic stimulation signals through the auditory pathway and influences the functional state of neural networks by utilizing parameters such as the type of acoustic stimulation, frequency range, modulation method, modulation rhythm, sound intensity range, stimulation duration, left and right ear output relationship, and stimulation combination. With the development of wearable EEG devices, mobile terminals, remote follow-up systems, and medical device software, acoustic stimulation neuromodulation is gradually becoming capable of long-term, continuous, home-based, and remote applications.

[0003] Current acoustic stimulation systems primarily focus on parameter settings and real-time feedback during single or short-term interventions, typically adjusting parameters based on current EEG signals, user feedback, or pre-defined rules. However, over long-term use, the same subject's responses to different acoustic stimulation parameters and signal types may vary with the number of uses, duration of use, changes in state, sleep patterns, emotional state, auditory tolerance, and therapeutic effects. Relying solely on single-instance feedback or current state variables for parameter adjustments fails to fully utilize the subject's accumulated historical data over a long period.

[0004] The existing technology has the following shortcomings:

[0005] First, existing acoustic stimulation systems typically lack structured organization of long-term historical EEG data and therapeutic feedback data, making it impossible to form an individual-oriented historical response dataset.

[0006] Second, existing systems do not focus on establishing the mapping relationship between acoustic stimulation parameters, acoustic stimulation signal types, neural state variables, and therapeutic feedback, making it difficult to explain the long-term effects of different parameter combinations and different acoustic stimulation signal types on the same subject.

[0007] Third, although some systems introduce machine learning or artificial intelligence models, they tend to directly use the model output as the result of automatic control, lacking physician confirmation, safety state machine constraints, and interpretability mechanisms, which is not conducive to risk management of medical device software.

[0008] Fourth, existing parameter optimization relies heavily on empirical rules or the current state, failing to fully utilize long-term trend analysis and making it difficult to achieve individualized shrinkage of the parameter search space.

[0009] Fifth, current efficacy assessments largely rely on subjective feedback, scales, or short-term observations, lacking objective EEG efficacy indicators based on historical EEG testing and analysis. This makes it difficult to systematically correlate changes in EEG characteristic values, EEG state variables, EEG rhythm stability, and brain network connectivity characteristics with efficacy trends. Summary of the Invention

[0010] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0011] A personalized optimization method for acoustic stimulation neuromodulation based on historical EEG data and therapeutic feedback is provided. This method establishes a parameter-signal-state-therapeutic effect mapping relationship based on historical EEG data, acoustic stimulation signal types, state variables, subjective feedback, therapeutic indicators, and objective EEG therapeutic indicators obtained during multiple acoustic stimulation interventions. This mapping relationship is then used to generate personalized recommendation results. The method includes the following steps:

[0012] S1: Obtain historical intervention data, which includes one or more of the following: EEG characteristics, acoustic stimulation parameters, acoustic stimulation signal type, EEG state variables, physiological indicators, subjective feedback, therapeutic efficacy indicators, objective EEG therapeutic efficacy indicators, usage duration, and safety events.

[0013] S2: Preprocess the historical intervention data to form an individual historical response dataset. The preprocessing includes time stamping, quality screening, anomaly marking, missing data handling, and structured organization.

[0014] S3: Based on the individual historical response dataset, establish a parameter-signal-state-efficacy mapping relationship between acoustic stimulation parameters, acoustic stimulation signal types, EEG state variables, EEG objective efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators;

[0015] S4: Based on the parameter-signal-state-therapeutic effect mapping relationship, construct an individual response model. The individual response model is used to describe the trend of changes in the subject's neurological state, therapeutic effect feedback, or risk under different acoustic stimulation parameters and different acoustic stimulation signal types.

[0016] S5: Generate individualized recommendation results based on the individual response model. The individualized recommendation results include individualized parameter recommendation range, acoustic stimulus signal type recommendation, parameter priority suggestion, parameter disabled range, parameter search space shrinkage result, efficacy trend prediction result or efficacy response probability.

[0017] S6: Output the individualized recommendation results to the doctor's terminal and the sound stimulation prescription template module or closed-loop control module;

[0018] S7: After confirmation by the doctor and under the constraints of the prescription template and / or the safety state machine, the individualized recommendation result is used for subsequent acoustic stimulus parameter initialization, acoustic stimulus signal type selection, parameter update suggestion generation, or parameter search space limitation.

[0019] S8: When insufficient EEG signal quality, abnormal fluctuations in state variables, poor user feedback, out-of-bounds parameters, increased safety risks, or abnormal system operation are detected, ignore the current individualized recommendation results and switch to conservative parameters, pause, stop, or rollback mechanisms under the constraints of the safety state machine.

[0020] S9: Record new historical intervention data during subsequent acoustic stimulation interventions and use it to update the individual's historical response dataset;

[0021] S10: Record the construction process of the individual response model, input data version, model version, changes in EEG objective efficacy indicators, feature contribution results, parameter recommendation results, doctor confirmation records, safety status switching records, and model update records.

[0022] In a preferred embodiment of the present invention, the EEG features include one or more of the following: power in different frequency bands, relative power, power stability, inter-band ratio, synchronicity, coherence, cross-frequency coupling, brain network connectivity features, rhythm stability, and event-related change features.

[0023] In a preferred embodiment of the present invention, the acoustic stimulation parameters include one or more of the following: acoustic stimulation type, frequency band range, one or more center frequencies, modulation method, modulation depth, modulation rhythm, sound intensity range, acoustic stimulation duration, acoustic stimulation energy, left and right ear output relationship, stimulation interval, and stimulation combination sequence; the acoustic stimulation signal type includes one or more of the following: pure tone, narrowband noise, broadband noise, modulated sound, rhythmic sound, natural sound, composite sound, musical sound, and pulse stimulation with a set waveform, wherein the modulated sound includes amplitude modulation sound, frequency modulation sound, or a mixture of amplitude modulation and frequency modulation sound.

[0024] In a preferred embodiment of the present invention, the physiological indicators include one or more of the following: heart rate, heart rate variability, respiratory-related indicators, sleep-related indicators, activity level, skin conductance response, blood oxygenation-related indicators, and body movement indicators; the subjective feedback includes user feedback on auditory stimulation tolerance, discomfort, symptom changes, sleep changes, attention status, fatigue level, emotional state, or user experience; the efficacy indicators include one or more of the following: changes in scale scores, degree of improvement in target state, degree of subjective improvement by the user, changes in sleep quality, changes in symptom burden, changes in compliance, physician follow-up evaluation, and objective efficacy indicators of EEG obtained based on historical EEG detection and analysis; the objective efficacy indicators of EEG include one or more of the following: changes in EEG characteristic values, changes in EEG state variables, changes in EEG rhythm stability, changes in brain network connectivity characteristics, parameter-EEG-efficacy correlation strength, efficacy trend prediction results, and efficacy response probability.

[0025] In a preferred embodiment of the present invention, the parameter-signal-state-therapeutic effect mapping relationship is represented by one or more of the following: rule relationship, statistical relationship, weighted scoring relationship, trend relationship, correlation matrix, response curve, or interpretable model parameters.

[0026] In a preferred embodiment of the present invention, the individual response model is one or more of the following: a rule-based model, a statistical model, a machine learning model, a rule-learning hybrid model or a multi-model fusion model, an interpretable scoring model or a trend prediction model; wherein, the machine learning model, rule-learning hybrid model or multi-model fusion model is used to train, fit, predict trends, score risks or analyze features on historical EEG features, acoustic stimulation parameters, acoustic stimulation signal types, state variables, objective EEG efficacy indicators, subjective feedback, physiological indicators and efficacy indicators.

[0027] In a preferred embodiment of the present invention, the individualized recommendation result includes one or more of the following: recommended parameter range, recommended parameter level, recommended parameter combination, recommended acoustic stimulus signal type, prohibited parameter range, parameter update priority, efficacy trend prediction result, efficacy response probability, and parameter search space boundary.

[0028] In a preferred embodiment of the present invention, the individual response model does not directly output disease diagnosis conclusions, nor is it directly used as the final treatment decision; the output results of the individual response model are used to assist doctors in confirmation, prescription template adjustment, or parameter updates under safety state machine constraints.

[0029] In a preferred embodiment of the present invention, the security state machine constraints include one or more of the following: signal quality gating, security threshold verification, disable condition judgment, observation period constraints, rollback rules, pause rules, and stop rules.

[0030] A personalized optimization system for acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback, comprising:

[0031] The historical data acquisition module is used to collect or record historical intervention data;

[0032] The data processing module is used to perform time stamping, quality screening, anomaly marking, missing data handling, and structured processing on historical intervention data to form individual historical response datasets;

[0033] The mapping relationship construction module is used to establish parameter-signal-state-efficacy mapping relationships between acoustic stimulation parameters, acoustic stimulation signal types, EEG state variables, EEG objective efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators.

[0034] The individual response model building module is used to build an individual response model based on the individual's historical response dataset and the parameter-signal-state-treatment mapping relationship;

[0035] The AI-assisted modeling and interpretation module is used to train, fit, predict trends, score risks, and analyze feature contributions of EEG features, acoustic stimulation parameters, acoustic stimulation signal types, state variables, objective EEG efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators in an individual's historical response dataset based on machine learning models, rule-learning hybrid models, or multi-model fusion models, and output interpretable auxiliary recommendation results.

[0036] The parameter recommendation module is used to generate personalized recommendation results based on the individual response model; the doctor's confirmation module is used to display or review the personalized recommendation results and record the doctor's confirmation, adjustment, locking or rejection information.

[0037] The safety constraint module is used to ignore the current individualized recommendation results and switch to conservative parameters, pause, stop, or rollback mechanisms when an abnormal state or safety risk is detected.

[0038] The model update and log module is used to record the model building process, input data version, model version, changes in EEG objective efficacy indicators, feature contribution results, parameter recommendation results, doctor confirmation records, safety status switching records, and model update records.

[0039] The beneficial effects of this invention are: by comprehensively collecting historical intervention data to integrate and generate individual historical response datasets, an individual response model is constructed, thereby autonomously generating individualized recommendation results, which improves the individual adaptability, security, and interpretability of the recommendation results. Moreover, by recording information at each stage or operation, the traceability, auditability, and quality management capabilities of the system are improved. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein:

[0041] Figure 1 This is a flowchart illustrating an individualized optimization method for acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback, according to the present invention.

[0042] Figure 2 This is a schematic diagram of the data structure of the individual historical response dataset in this invention;

[0043] Figure 3 This is a schematic diagram illustrating the parameter-signal-state-therapeutic effect mapping relationship and the construction of the individual response model in this invention;

[0044] Figure 4 This is a schematic diagram illustrating the relationship between individualized recommendation, safety constraints, and model updates in this invention. Detailed Implementation

[0045] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] In practical applications, many medical systems record historical intervention data from multiple acoustic stimulation interventions. This historical information contains individualized response patterns of subjects to different acoustic stimulation protocols. This application constructs individual response relationships between parameters, signals, states, and therapeutic effects based on this historical data, and generates objective EEG efficacy indicators. This aims to narrow the parameter search space, reduce ineffective parameter attempts, and improve individual adaptation efficiency during subsequent interventions, while also providing supplementary information for physician prescription adjustments. Simultaneously, by combining AI-assisted modeling and interpretive analysis, it generates individualized parameter recommendation ranges, efficacy trend predictions, or efficacy response probabilities. These are then confirmed by the physician, constrained by prescription templates, and subject to safety state machine constraints for subsequent closed-loop control or prescription template invocation. This improves the individual adaptability, safety, interpretability, and traceability of long-term acoustic stimulation neuromodulation processes.

[0047] The embodiments of the present invention include:

[0048] A personalized optimization method for acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback, comprising the following steps:

[0049] Step 1: During multiple acoustic stimulation interventions, acquire the subject's historical intervention data, which includes one or more of the following: historical EEG characteristics, historical acoustic stimulation parameters, historical acoustic stimulation signal types, historical state variables (including the patient's described state and corresponding EEG state variables), historical physiological indicators, historical subjective feedback, historical efficacy indicators, historical EEG objective efficacy indicators, usage duration, and safety events.

[0050] Furthermore, EEG characteristics include one or more of the following: power in different frequency bands, relative power, power stability, inter-band ratio, synchronicity, coherence, cross-frequency coupling, brain network connectivity characteristics, rhythm stability, and event-related change characteristics.

[0051] Furthermore, the acoustic stimulation parameters include one or more of the following: acoustic stimulation type, frequency band range, one or more center frequencies, modulation method, modulation depth, modulation rhythm, acoustic intensity range, acoustic stimulation duration, acoustic stimulation energy, left and right ear output relationship, stimulation interval, and stimulation combination sequence.

[0052] Furthermore, the types of acoustic stimulus signals include one or more of the following: pure tone, narrowband noise, broadband noise, modulated sound, rhythmic sound, natural sound, complex sound, musical sound, and pulse stimulation with a set waveform.

[0053] In some embodiments of this application, the modulated sound includes amplitude modulation (AM), frequency modulation (FM), or a mixture of AM and FM.

[0054] Furthermore, physiological indicators include one or more of the following: heart rate, heart rate variability, respiratory-related indicators, sleep-related indicators, activity level, skin conductance, blood oxygenation-related indicators, and body movement indicators.

[0055] Furthermore, subjective feedback includes user feedback on their tolerance to auditory stimulation, discomfort, changes in symptoms, changes in sleep, attention span, fatigue level, emotional state, or user experience.

[0056] Furthermore, efficacy indicators include one or more of the following: changes in scale scores, degree of improvement in target state, degree of subjective improvement by the user, changes in sleep quality, changes in symptom burden, changes in compliance, physician follow-up evaluation, and objective efficacy indicators of EEG based on historical EEG detection and analysis.

[0057] Furthermore, objective efficacy indicators of EEG include one or more of the following: changes in EEG characteristic values, changes in EEG state variables, changes in EEG rhythm stability, changes in brain network connectivity characteristics, parameter-EEG-efficacy correlation strength, efficacy trend prediction results, and efficacy response probability.

[0058] Step 2: Time-stamp, quality-filter, anomaly-marking, missing data handling, and structure-organizing the historical intervention data to form an individual historical response dataset.

[0059] Step 3: Based on the individual historical response dataset, establish the parameter-signal-state-therapeutic effect mapping relationship among acoustic stimulation parameters, acoustic stimulation signal type, EEG state variables, EEG objective efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators.

[0060] Step 4: Based on the parameter-signal-state-therapeutic effect mapping relationship, construct an individual response model to describe the trend of changes in the subject's neural state, therapeutic effect feedback, or risk under different acoustic stimulation parameters and different acoustic stimulation signal types.

[0061] Furthermore, the individual response model is one or more of the following: rule-based model, statistical model, machine learning model, rule-learning hybrid model, multi-model fusion model, interpretable scoring model, or trend prediction model.

[0062] Step 5: Based on the individual response model, generate individualized recommendation results, including individualized parameter recommendation range, acoustic stimulus signal type recommendation, parameter priority suggestion, parameter prohibition range, parameter search space contraction results, efficacy trend prediction results or efficacy response probability.

[0063] Furthermore, the individual response model does not directly output disease diagnosis conclusions, nor does it directly serve as the final treatment decision; its output results are used to assist doctors in confirmation, prescription template adjustment, or parameter updates under safety state machine constraints.

[0064] Furthermore, the scope of individualized parameter recommendations includes one or more of the following: recommended parameter range, recommended parameter level, recommended parameter combination, recommended acoustic stimulus signal type, prohibited parameter range, parameter update priority, efficacy trend prediction result, efficacy response probability, and parameter search space boundary.

[0065] Step Six: Output the recommended range of individualized parameters, recommended acoustic stimulation signal types, suggested parameter priorities, prohibited parameter ranges, parameter search space contraction results, efficacy trend prediction results, or efficacy response probability to the doctor for confirmation, and then send them to the acoustic stimulation prescription template module or closed-loop control module for safety status judgment.

[0066] Step 7: On the doctor's end, confirm and, under the constraints of the prescription template and / or the safety state machine, use the individualized recommendation results output by the individual response model for subsequent acoustic stimulus parameter initialization, acoustic stimulus signal type selection, parameter update suggestion generation, or parameter search space limitation. Further, the safety state machine constraints include one or more of the following: signal quality gating, safety threshold verification, disable condition judgment, observation period constraint, rollback rule, pause rule, and stop rule.

[0067] Step 8: When using the current personalized recommendation results, if an anomaly is detected (insufficient EEG signal quality, abnormal fluctuations in state variables, poor user feedback, parameter out-of-bounds, increased safety risks, or abnormal system operation), the current personalized recommendation results are ignored, and the system switches to a conservative parameter, pause, stop, or rollback mechanism under safety constraints.

[0068] Step 9: Continuously record new EEG characteristics, sound stimulation parameters, sound stimulation signal types, state variables, subjective feedback, physiological indicators, efficacy indicators, and objective EEG efficacy indicators during subsequent sound stimulation interventions, and use them to update the individual's historical response dataset.

[0069] Step 10: Record the construction process of the individual response model, the version of the input data, the model version, changes in the objective efficacy indicators of EEG, the feature contribution results, the parameter recommendation results, the doctor's confirmation record, the safety status switching record, and the model update record.

[0070] A personalized optimization system for acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback includes:

[0071] The historical data acquisition module is used to record historical intervention data during multiple acoustic stimulation interventions.

[0072] The data processing module is used to perform time stamping, quality screening, anomaly marking, missing data handling, and structured processing on the recorded historical intervention data to form an individual historical response dataset;

[0073] The mapping relationship construction module is used to establish parameter-signal-state-efficacy mapping relationships between acoustic stimulation parameters, acoustic stimulation signal types, EEG state variables, EEG objective efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators.

[0074] The individual response model building module is used to build individual response models based on individual historical response datasets and parameter-signal-state-treatment mapping relationships;

[0075] The AI-assisted modeling and interpretation module is used to train, fit, predict trends, score risks, and analyze feature contributions of EEG features, acoustic stimulation parameters, acoustic stimulation signal types, state variables, objective EEG efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators in an individual's historical response dataset based on machine learning models, rule-learning hybrid models, or multi-model fusion models, and output interpretable auxiliary recommendation results.

[0076] The parameter recommendation module is used to generate personalized recommendation results based on the individual response model;

[0077] The doctor-side confirmation module is used to display or review the results output by the parameter recommendation module (individualized recommendation results), and record doctor-side confirmation, adjustment, locking, or rejection information;

[0078] The safety constraint module is used to ignore the output of the individual response model and switch to conservative parameters, pause, stop or rollback mechanism when an abnormal state or safety risk is detected.

[0079] The model update and log module is used to record the model building process, input data version, model version, changes in EEG objective efficacy indicators, feature contribution results, parameter recommendation results, doctor confirmation records, safety status switching records, and model update records.

[0080] like Figure 1 As shown, the process of this application is based on historical EEG data, sound stimulation parameters, sound stimulation signal types, state variables, physiological indicators, subjective feedback, efficacy indicators, objective EEG efficacy indicators and safety events obtained during multiple sound stimulation interventions. It establishes a parameter-signal-state-efficacy mapping relationship, constructs an individual response model, and, under the constraints of doctor confirmation, prescription template constraints and safety state machine constraints, uses the model output results for subsequent sound stimulation parameter initialization, parameter update or parameter search space limitation.

[0081] Starting with step 101, the system acquires historical intervention data from multiple interventions. This intervention data includes EEG characteristics, acoustic stimulation parameters, acoustic stimulation signal types, state variables, physiological indicators, subjective feedback, efficacy indicators, objective EEG efficacy indicators, and safety events generated by the subjects during multiple acoustic stimulation interventions, observation periods, or follow-up cycles. This provides a historical data foundation for long-term individualized optimization.

[0082] In step 102, the system performs data processing and quality control on historical intervention data from multiple interventions. This processing includes time stamping, quality filtering, anomaly labeling, and missing data handling. This step removes low-quality data, identifies anomalous events, and organizes data from different intervention periods into structured data that can be used for subsequent modeling and retrospective analysis.

[0083] Step 103 generates an individual historical response dataset. This dataset is used to store changes in the subject's EEG state, subjective feedback, physiological indicators, therapeutic efficacy indicators, objective EEG therapeutic efficacy indicators, and safety event records under different acoustic stimulation parameters, different acoustic stimulation signal types, and different usage scenarios.

[0084] In step 104, the system establishes a parameter-signal-state-therapeutic effect mapping relationship based on the individual's historical response dataset. This mapping relationship describes the correlation between acoustic stimulation parameters, acoustic stimulation signal types, EEG characteristics, state variables, physiological indicators, subjective feedback, therapeutic effect indicators, and objective EEG therapeutic effect indicators. Through this mapping relationship, the system can analyze the impact of different acoustic stimulation parameters and acoustic stimulation signal types on changes in the subject's neurological state, therapeutic effect, and risk.

[0085] Step 105 involves constructing an individual response model. This model describes the trends in the subject's neurological state, therapeutic feedback, or risk under different acoustic stimulation parameters and signal types, and can be used to generate subsequent parameter recommendations. This model does not directly output a disease diagnosis, nor is it directly used as the final treatment decision.

[0086] In step 106, the system generates individualized recommendation results based on the individual response model. These individualized recommendations may include recommended parameter ranges, recommended acoustic stimulus signal types, parameter priority suggestions, prohibited parameter ranges, parameter search space contraction results, efficacy trend prediction results, and efficacy response probabilities. This step aims to reduce ineffective parameter attempts and improve individual adaptation efficiency in subsequent acoustic stimulus interventions.

[0087] Step 107 indicates that the doctor reviews or confirms the recommendations. The personalized recommendations output by the individual response model can be displayed on the doctor's end, where they can be reviewed, confirmed, adjusted, locked, or rejected. This step emphasizes that the model output serves only as an auxiliary recommendation, not as an automatic substitute for the doctor's final decision.

[0088] In step 108, the individualized recommendation results, confirmed by the doctor, are used for subsequent acoustic stimulus parameter initialization, parameter updates, or parameter search space constraints under the constraints of the prescription template and the safety state machine. In other words, the model output results need to be jointly constrained by the prescription template, safety threshold, disabling conditions, observation period rules, backoff rules, and the safety state machine before they can be used for subsequent closed-loop acoustic stimulus parameter adjustments.

[0089] In step 109, the system determines whether any abnormalities have occurred. Abnormalities may include insufficient EEG signal quality, abnormal fluctuations in state variables, poor user feedback, parameters exceeding limits, increased safety risks, deterioration of objective EEG efficacy indicators, or abnormal system operation.

[0090] If no abnormality is found in step 109, the process proceeds to step 110. In step 110, during acoustic stimulation intervention, the system continuously records new EEG characteristics, acoustic stimulation parameters, acoustic stimulation signal types, state variables, physiological indicators, subjective feedback, efficacy indicators, objective EEG efficacy indicators, and safety events. The newly added data can be used to subsequently update the individual's historical response dataset and individual response model.

[0091] If an anomaly is detected in step 109, the process proceeds to step 111. In step 111, the system ignores the individual response model's recommendations and switches to a conservative parameter, pause, stop, or rollback mechanism under a safe state machine constraint. This step is used to prevent model recommendations from continuing to affect the acoustic stimulus output or parameter updates under abnormal conditions, thereby improving system safety.

[0092] Step 112 represents model updates and version recording. The system records data version, model version, feature contribution results, parameter recommendation results, doctor confirmation records, safety status switching records, and changes in objective EEG efficacy indicators. When new data meets the preset model update conditions, the system can update the individual response model; the results generated after the model update still need to be confirmed by the doctor and subject to safety constraints before they can be used for subsequent parameter optimization.

[0093] In summary, the process outlined in this application forms a long-term individualized optimization chain: "multiple intervention data acquisition—data organization and quality processing—construction of individual historical response datasets—establishment of parameter-signal-state-therapeutic effect mapping relationships—construction of individual response models—personalized parameter recommendation—physician review and confirmation—invocation of prescription templates and safety state machine constraints—anomaly risk assessment—safety rollback—addition of data records—model updates and version tracking." Through this process, this invention can utilize historical EEG data, objective EEG efficacy indicators, and efficacy feedback data, combined with AI or multi-model fusion analysis, to form an interpretable, traceable, and safety-constrained individualized optimization scheme for acoustic stimulation neuromodulation.

[0094] like Figure 2 As shown, this application presents the data structure, data sources, data quality processing, data version tracing, and subsequent application relationships of the individual historical response dataset. This illustrates how this application structures and organizes multi-source data generated during multiple acoustic stimulation interventions, observation periods, or follow-ups to form a data foundation that can be used to construct parameter-signal-state-therapeutic effect mapping relationships, build individual response models, and recommend individualized parameters.

[0095] In the diagram, solid arrows represent data flow or call relationships, while dashed arrows represent constraints, updates, or record relationships.

[0096] The Individual Historical Response Dataset 201 serves as the core data source for long-term individualized optimization. This dataset is formed by accumulating historical data from subjects across multiple acoustic stimulation interventions, observation periods, or follow-up periods. It is used to preserve historical intervention data such as acoustic stimulation parameters, acoustic stimulation signal types, EEG characteristics, state variables, physiological indicators, subjective feedback, efficacy indicators, safety events, and time quality markers.

[0097] Individual historical response dataset 201 includes the following data fields:

[0098] (a) Subject identification and scenario information 202: Used to record subject identification, usage scenario, usage environment, intervention stage, device mode, doctor end or home end usage information, etc., in order to distinguish different subjects, different application scenarios and different intervention stages.

[0099] (b) Intervention Session Index 203: Used to number and index each acoustic stimulus intervention, each observation period, or each follow-up period, so that the system can track changes in individual responses in chronological order or at each stage of the intervention.

[0100] (c) EEG Feature Field 204: Used to record EEG features extracted during each intervention or observation period. EEG features may include one or more of the following: power in different frequency bands, relative power, power stability, inter-band ratio, synchronicity, coherence, cross-frequency coupling, brain network connectivity features, rhythm stability, and event-related change features.

[0101] (d) Sound Stimulation Parameter Field 205: Used to record the sound stimulation parameters used in each intervention. Sound stimulation parameters may include sound stimulation type, frequency range, center frequency, modulation method, modulation depth, modulation rhythm, sound intensity range, sound stimulation duration, sound stimulation energy, left and right ear output relationship, stimulation interval, and stimulation combination sequence, etc.

[0102] (e) Acoustic Stimulus Signal Type Field 206: This field records the specific acoustic stimulus signal morphology, such as pure tone, narrowband noise, broadband noise, modulated sound, rhythmic sound, natural sound, complex sound, musical sound, or pulse stimulation with a set waveform. This field is used to distinguish the effects of different acoustic stimulus signal types on changes in the subject's neural state, subjective feedback, and therapeutic efficacy indicators.

[0103] (f) State Variables / Physiological Indicators Field 207: Used to record the state variables generated and the physiological indicators collected synchronously during each intervention. State variables may include network stability, cortical synchronization status, arousal level, target state deviation, state change trend or risk status; the physiological indicators may include heart rate, heart rate variability, respiratory-related indicators, sleep-related indicators, activity level, skin conductance response, blood oxygenation-related indicators or body movement indicators, etc.

[0104] (g) Subjective Feedback / Efficacy Indicator Field 208: Used to record user subjective feedback and efficacy-related indicators.

[0105] Subjective feedback may include tolerance to sound stimulation, discomfort, changes in symptoms, changes in sleep, attention, fatigue, emotional state, or user experience; efficacy indicators may include changes in scale scores, degree of improvement in target state, degree of subjective improvement by the user, changes in sleep quality, changes in symptom burden, changes in compliance, physician follow-up evaluation, and objective efficacy indicators based on historical EEG detection and analysis.

[0106] Objective efficacy indicators of EEG can include one or more of the following: changes in EEG characteristic values, changes in EEG state variables, changes in EEG rhythm stability, changes in brain network connectivity characteristics, the strength of the correlation between parameters, EEG and efficacy, efficacy trend prediction results, and efficacy response probability.

[0107] (h) Time stamp / quality stamp / anomaly stamp field 209: Used to record the time information, data quality information and anomaly event stamps for each historical response data.

[0108] Time stamps are used to determine the position of data before, during, after, during, or in the observation or follow-up period; quality stamps are used to identify whether the data meets the requirements for subsequent modeling; anomaly stamps are used to record situations such as insufficient signal quality, unpleasant user feedback, out-of-bounds parameters, increased safety risks, or abnormal system operation.

[0109] The data sources for Individual Historical Response Dataset 201 mainly include:

[0110] • Intervention session record 210: Provides basic index information for each intervention, observation period, or follow-up cycle;

[0111] • EEG Feature Recording 211: Provides EEG signal detection and analysis results;

[0112] • Acoustic stimulus parameters / signal recording 212: Provides information on acoustic stimulus parameters and acoustic stimulus signal type;

[0113] • State variable / physiological index record 213: provides information on neural state and physiological feedback;

[0114] • Subjective Feedback / Therapeutic Indicator Records 214: Provides records related to user subjective feedback, therapeutic indicators, and objective EEG therapeutic indicators;

[0115] • Security Incident / Follow-up Record 215: Provides information on security incidents, abnormal events, and doctor follow-up.

[0116] The above multi-source data are processed and then entered into the individual historical response dataset 201.

[0117] Data quality processing and structuring involve time stamping, quality screening, missing value handling, anomaly labeling, and structuring of raw historical data to improve dataset consistency, usability, and modeling reliability. This includes labeling or processing low-quality EEG data, missing feedback, anomalous events, or incomplete records to prevent them from directly impacting model construction and parameter recommendation.

[0118] Subsequent applications of the Individual Historical Response Dataset 201 include:

[0119] • It can be used for parameter-signal-state-therapeutic effect mapping216, that is, to establish the mapping relationship between acoustic stimulation parameters, acoustic stimulation signal type, EEG state variables, EEG objective therapeutic effect indicators, subjective feedback, physiological indicators and therapeutic effect indicators.

[0120] • It can be used to build individual response models217, and fit individual response patterns through rule models, statistical models, machine learning models, rule-learning hybrid models or multi-model fusion models.

[0121] Furthermore, the system can perform parameter recommendation / search space shrinking 218 based on the output of the individual response model to generate personalized recommendation results. These results can then be reviewed and applied by the doctor 219, where they can be confirmed, adjusted, locked, or rejected.

[0122] Furthermore, through data version and traceability record 221, the data version number, change record, data source, and traceability information are recorded. Through this module, the system can trace the corresponding data source, processing process, and version changes when the model is updated, parameter recommendation results change, doctor confirmation occurs, or a security event occurs.

[0123] In summary, this application establishes a data chain encompassing "multi-source historical data input—data quality processing and structured organization—construction of individual historical response datasets—parameter-signal-state-therapeutic effect mapping—construction of individual response models—parameter recommendation and search space contraction—physician-side review and application." Through this data structure, this application can transform historical EEG data, acoustic stimulation parameters, acoustic stimulation signal types, objective EEG therapeutic effects indicators, subjective feedback, therapeutic effects indicators, and safety events generated during long-term acoustic stimulation intervention into a modelable, interpretable, and traceable foundation for individualized optimization.

[0124] like Figure 3As shown, this application establishes a parameter-signal-state-therapeutic effect mapping relationship based on an individual historical response dataset, and further constructs an individual response model. This illustrates how this application integrates historical acoustic stimulation parameters, acoustic stimulation signal types, EEG characteristics, neural state variables, therapeutic effect feedback, objective EEG therapeutic effect indicators, and safety events into an interpretable individual response relationship, and generates individualized parameter recommendations, risk warnings, and model update results based on this relationship.

[0125] In the diagram, solid arrows represent the main process or data flow, dashed arrows represent feedback, update, or constraint relationships, and dashed boxes represent functional groups or model composition.

[0126] (1) Obtain individual historical response dataset 301. This dataset consists of historical data obtained during multiple acoustic stimulation interventions, observation periods, and follow-ups of the subjects. It is the data foundation for establishing mapping relationships and constructing individual response models. The specific data obtained includes the following:

[0127] • Acoustic stimulation parameter feature 302 describes the adjustable controllable conditions of acoustic stimulation output, including acoustic stimulation type, frequency range, center frequency, modulation mode, sound intensity range, stimulation duration, stimulation rhythm, stimulation interval, and acoustic stimulation energy. This type of feature is used to analyze the impact of different parameter settings on changes in neural state and therapeutic feedback.

[0128] • Acoustic stimulus signal type feature 303 describes the specific morphology of the acoustic stimulus signal, including pure tone, narrowband noise, broadband noise, modulated sound, rhythmic sound, natural sound, complex sound, musical sound, and impulse sound. This type of feature is used to distinguish the effects of different acoustic signal morphologies on the subject's EEG state, subjective feedback, and therapeutic trend.

[0129] • Neural state and EEG characteristics 304 are used to characterize the neural network state of subjects during acoustic stimulation intervention, including frequency band power, synchronicity, coherence, cross-frequency coupling, brain network connectivity characteristics, and state variables. These characteristics are used to reflect the direction and degree of change in neural state after acoustic stimulation.

[0130] • Therapeutic efficacy and feedback feature 305 is used to describe the feedback results and efficacy-related information after acoustic stimulation intervention, including subjective feedback, physiological indicators, efficacy indicators, objective EEG efficacy indicators, safety events, and follow-up records. The objective EEG efficacy indicators may include one or more of the following: changes in EEG characteristic values, changes in EEG state variables, changes in EEG rhythm stability, changes in brain network connectivity characteristics, parameter-EEG-efficacy correlation strength, efficacy trend prediction results, and efficacy response probability.

[0131] (2) Input the above features into the mapping relationship construction module 306 for parameters-signals-states-therapeutic effects. The mapping relationship construction module 306 is used to establish the correlation between acoustic stimulation parameters, acoustic stimulation signal types, neural state variables, EEG characteristics, subjective feedback, physiological indicators, therapeutic effect indicators and objective EEG therapeutic effect indicators.

[0132] (2.1) Association analysis / rule extraction 3061 is used to extract the correlation, rule relationship or constraint relationship between different parameters, different signal types, different EEG state changes and different efficacy feedback from historical intervention data.

[0133] Furthermore, this step can employ statistical association, rule mining, mutual information analysis, causal analysis, or other interpretable analysis methods.

[0134] (2.2) Response Curve / Association Matrix Generation 3062: This step generates response curves corresponding to different parameters or acoustic stimulus signal types, and forms correlation matrices such as parameter-state, signal-state, parameter-therapeutic effect, signal-therapeutic effect, and state-therapeutic effect. This step helps to show the correspondence between a certain acoustic stimulus parameter or acoustic stimulus signal type and changes in EEG state, subjective feedback, or therapeutic effect indicators.

[0135] (2.3) Trend and risk relationship modeling 3063 is used to analyze the relationship between specific parameters, acoustic stimulation signal types or parameter combinations and long-term efficacy trends, stable state trends, discomfort feedback or safety events, and to form the basis for risk assessment, trend prediction or safety boundary judgment.

[0136] (3) Input the mapping relationship construction results into the individual response model construction module 307. Based on the individual historical response dataset 301 and the output of the mapping relationship construction module 306, form an individual response model exclusive to the subject, which is used to describe the subject's neural state change trend, therapeutic effect feedback trend and risk change trend under different acoustic stimulation parameters and acoustic stimulation signal types.

[0137] Furthermore, the individual response model building module 307 can include various model formats:

[0138] • Rule model 3071 is used to express a rule base formed based on expert knowledge, security constraints, prohibition conditions, and conservative parameter strategies;

[0139] • Statistical model 3072 is used to establish statistical relationships between parameters, signals, states, and therapeutic effects;

[0140] • Machine learning model 3073 was used to fit the nonlinear relationship between historical EEG features, acoustic stimulation parameters, acoustic stimulation signal type, state variables, objective EEG efficacy indicators, and efficacy feedback.

[0141] • Rule-learning hybrid model 3074 is used to combine manually set safety rules with machine learning fitting results;

[0142] The multi-model fusion 3075 is used to integrate the results of multiple models through weighted fusion, stacking, voting or Bayesian fusion, etc., to improve the stability and interpretability of recommendations.

[0143] Furthermore, the individual response model construction module 307 also includes an AI-assisted modeling and interpretation module 307A, used for further analysis of individual historical response data and model output. This module can perform feature contribution analysis, variable importance analysis, risk scoring, long-term trend prediction, and efficacy response probability assessment to generate interpretable auxiliary recommendation results. The output of this unit is not directly used as the final treatment decision, nor does it directly output disease diagnosis conclusions. Instead, it is used to assist doctors in confirmation, prescription template adjustment, or parameter updates under safety state machine constraints.

[0144] (4) Based on the output of the individual response model construction module 307, generate personalized recommendation results, the specific contents of which include:

[0145] • The recommended range of individualized parameters is 308, which is used to indicate the parameter range, parameter level, type of sound stimulus signal or combination of mode that is more suitable for a subject in subsequent sound stimulation intervention.

[0146] • Parameter priority suggestion / search space shrinkage result 309. This result is used to suggest the direction of subsequent priority exploration of parameters and to shrink the parameter search space in order to reduce invalid parameter attempts and improve the efficiency of subsequent intervention.

[0147] • When historical intervention data indicates that certain parameters, acoustic stimulation signal types, or parameter combinations are associated with discomfort feedback, worsening condition, decreased objective EEG efficacy indicators, or safety events, the system can generate a disabled parameter range / risk warning 310 to indicate high-risk combinations, low-priority ranges, or precautions.

[0148] (5) Individualized recommendation results are entered into the doctor's end for confirmation / prescription template or closed-loop control call 311.

[0149] (5.1) The doctor can review, confirm, adjust, lock or reject the model output results; the results confirmed by the doctor can be used to adjust the prescription template, or called by the closed-loop control module under safety constraints.

[0150] (5.2) Safety state constraint 312 is used to ensure that the model output will not directly affect the acoustic stimulus output under abnormal conditions. When an abnormality or increased risk is detected, the system can ignore the model suggestion and trigger conservative parameters, pause, stop or rollback mechanisms.

[0151] (6) Subsequent intervention results and new data 313: During subsequent acoustic stimulation interventions, the system continues to record new acoustic stimulation parameters, acoustic stimulation signal types, EEG characteristics, state variables, physiological indicators, subjective feedback, efficacy indicators, objective EEG efficacy indicators, and safety events. New data can be fed back to the individual historical response dataset 301, the mapping relationship construction module 306, and the individual response model construction module 307 for continuous model optimization.

[0152] (8) Model Update and Version Records 314: Record model updates, version management, data versions, change records, traceability information, and audit logs. When new data meets the preset quality and quantity conditions, the system can update the individual response model; if the recommendation results change too much or the security risk increases after the model update, it can revert to the previous stable model version.

[0153] In summary, this application establishes a technical chain encompassing "individual historical response dataset—acoustic stimulus parameter features—acoustic stimulus signal type features—neural state and EEG features—therapeutic efficacy and feedback features—construction of parameter-signal-state-therapeutic efficacy mapping relationship—individual response model construction—individualized parameter recommendation—physician confirmation—safety constraints—subsequent data feedback—model version update." This application can utilize long-term historical EEG data and therapeutic efficacy feedback data, combined with AI-assisted analysis, to form an interpretable, traceable, and safety-constrained individualized acoustic stimulus parameter optimization scheme.

[0154] The relationships between individual response model output, doctor confirmation, prescription template or closed-loop control call, safety constraints, exception handling, subsequent data aggregation, and model updates are as follows: Figure 4 As shown, the individualized recommendation results generated by the individual response model in this application are not directly used as the final treatment decision or automatic control instruction. Instead, under the constraints of doctor confirmation, prescription template constraints, and safety state machine constraints, they are used for subsequent acoustic stimulus parameter initialization, parameter updates, and parameter search space constraints.

[0155] Solid arrows represent the main flow or call flow, while dashed arrows represent constraint flow, update flow, or record flow.

[0156] The specific steps in this stage include:

[0157] Step 401: The individual response model outputs individualized recommendation results, which may include the recommended parameter range, parameter priority, disabled parameter range, and risk warnings.

[0158] Among them, the recommended parameter range is used to indicate the parameter range that can be preferentially selected for subsequent acoustic stimulation parameters; the parameter priority is used to indicate the parameter update order; the disabled parameter range is used to mark the parameter area that is not suitable for continued attempt or has a high risk; and the risk warning is used to indicate parameters or signal combinations that may be related to user discomfort, deterioration of condition, safety events or adverse changes in objective EEG efficacy indicators.

[0159] Step 402 involves confirmation or adjustment on the doctor's end. The individualized recommendation results output by the individual response model can be sent and displayed on the doctor's end, where they can be reviewed, confirmed, adjusted, locked, or rejected. This step emphasizes that the model output serves only as an auxiliary recommendation result, rather than directly replacing the doctor's final decision.

[0160] Step 403 involves calling the prescription template or closed-loop control module. The recommended results, confirmed or adjusted by the doctor, can be called by the system or an external acoustic stimulation prescription template module or closed-loop control module for subsequent acoustic stimulation parameter initialization, parameter update suggestion generation, or parameter search space limitation.

[0161] Step 404 involves initializing, updating, and limiting the search space for subsequent acoustic stimuli. In this step, the system initializes or adjusts the parameters for subsequent acoustic stimuli based on the confirmed individualized recommendation results, and limits the parameter search space to reduce invalid parameter attempts and overexploration.

[0162] Security constraints 405 include security threshold verification 4051, disable condition judgment 4052, observation period constraints 4053, and rollback / pause / stop rules 4054.

[0163] Safety threshold check 4051 is used to determine whether the parameter exceeds the preset boundary; disable condition check 4052 is used to determine whether the current state prohibits the application of model suggestions; observation period constraint 4053 is used to restrict the continuous adjustment behavior after parameter update; rollback / pause / stop rule 4054 is used to switch to a safer operating strategy when the abnormal state or risk increases.

[0164] In step 406, the system determines whether any abnormalities or increased risks have occurred:

[0165] If no abnormalities or increased risk are observed, the process proceeds to step 407: The system continuously records new data during the acoustic stimulation intervention. This new data may include EEG characteristics, acoustic stimulation parameters, acoustic stimulation signal type, state variables, physiological indicators, subjective feedback, efficacy indicators, objective EEG efficacy indicators, and safety events during the subsequent intervention process.

[0166] If an anomaly or increased risk occurs, the process proceeds to step 408: the system ignores the individualized recommendation results and switches to the safety mechanism. In other words, under abnormal or risky conditions, even if the individual response model has already generated a recommendation result, the system will not continue to update parameters according to that recommendation result.

[0167] Step 409: Implement conservative parameter settings, pause, stop, or rollback mechanisms. The system can, based on safety state machine constraints, switch to conservative parameter configuration, pause closed-loop parameter updates, stop acoustic stimulus output, or roll back to the most recent stable parameters or the parameter configuration corresponding to a previous stable model version.

[0168] Step 410, Summary of Subsequent Results and New Data: Regardless of whether the system continues to intervene or enters the safety mechanism, subsequent changes in state, user feedback, parameter results, safety events, and changes in objective EEG efficacy indicators can be summarized in this step to provide a basis for subsequent model updates.

[0169] Step 411: Determine whether the model update conditions are met. Model update conditions may include: the amount of new data reaching a preset quantity, data quality meeting requirements (meeting clinical practice requirements and EEG analysis requirements), completion of the observation period, confirmation from the doctor, changes in model performance meeting preset conditions, changes in objective EEG efficacy indicators meeting preset analysis conditions, or completion of traceability for safety events, etc.

[0170] If the model update conditions are not met, the system will not update the model, but will continue to run based on the current model version or the current security policy.

[0171] If the model update conditions are met, proceed to step 412.

[0172] Step 412, Model Update and Version Recording. The system updates the individual response model based on the new data and records the model version, data version, update time, changes in recommendation results, feature contribution results, changes in EEG objective efficacy indicators, and the basis for the update. After the model is updated, individualized recommendation results can be regenerated, but the new model output still needs to be confirmed by the doctor and subject to safety constraints before it can be applied.

[0173] Step 413 indicates the recorded content. The system records information such as data version, model version, recommendation results, doctor confirmation records, and safety status switching records. This record is used to achieve traceable management of the entire process of model updates, parameter recommendations, doctor confirmation, and safety incident handling.

[0174] In summary, this application establishes a complete chain: "Individual response model output—Doctor confirmation—Prescription template or closed-loop control invocation—Subsequent parameter initialization and search space limitation—Safety constraint judgment—Ignoring model suggestions and entering the safety mechanism in case of anomalies—New data aggregation—Model update and version recording." Through this chain, this application can fully utilize individual response models or AI-assisted analysis results for parameter recommendation while ensuring that the model output is subject to doctor confirmation, prescription template constraints, and safety state machine constraints, thereby improving the safety, interpretability, and traceability of the long-term acoustic stimulation neuromodulation individualized optimization process.

[0175] Example 1: Construction of an Individual Historical Response Dataset

[0176] In some embodiments of this application, the system records data corresponding to each intervention during multiple acoustic stimulation interventions received by the subject. The data includes one or more of the following: electroencephalogram (EEG) characteristics, acoustic stimulation parameters, acoustic stimulation signal type, state variables, physiological indicators, subjective feedback, efficacy indicators, objective EEG efficacy indicators, usage duration, and safety events.

[0177] Electroencephalogram (EEG) characteristics can include power across different frequency bands, relative power, power stability, inter-band ratios, synchronicity, coherence, cross-frequency coupling, brain network connectivity characteristics, rhythm stability, and event-related changes. Acoustic stimulation parameters can include stimulus type, frequency range, center frequency, modulation method, modulation depth, modulation rhythm, sound intensity range, stimulus duration, stimulus energy, left-right ear output relationship, stimulus interval, and stimulus combination sequence. Acoustic stimulus signal types can include pure tones, narrowband noise, broadband noise, modulated sound, rhythmic sound, natural sound, complex sound, musical sound, and pulse stimulation with a set waveform.

[0178] Physiological indicators may include heart rate, heart rate variability, respiratory-related indicators, sleep-related indicators, activity level, skin conductance, blood oxygenation-related indicators, and body movement indicators. Subjective feedback may include user feedback on auditory stimulation tolerance, discomfort, symptom changes, sleep changes, attentional status, fatigue level, emotional state, or user experience. Therapeutic efficacy indicators may include changes in scale scores, degree of improvement in target state, user subjective improvement, changes in sleep quality, changes in symptom burden, changes in compliance, physician follow-up evaluation, and objective EEG efficacy indicators.

[0179] The system performs time stamping, quality screening, anomaly marking, missing data handling, and structured processing on the above data to form an individual historical response dataset. Each historical response record can correspond to one acoustic stimulus intervention process, one observation period, or one follow-up period.

[0180] Example 2: Generation of Objective Therapeutic Indicators for Electroencephalography

[0181] In some embodiments of this application, the system extracts changes in EEG characteristic values, changes in EEG state variables, changes in EEG rhythm stability, and changes in brain network connectivity characteristics before and after intervention based on multiple historical EEG detection and analysis results. It then performs correlation analysis with changes in scale scores, user subjective feedback, doctor follow-up evaluations, and safety event records to generate objective EEG efficacy indicators.

[0182] Objective EEG efficacy indicators are used to characterize changes in the objective neural state of subjects under different acoustic stimulation parameters and different acoustic stimulation signal types. For example, the system can calculate the degree of improvement in EEG rhythm stability, the trend of changes in brain network connectivity characteristics, or the strength of the parameter-EEG-efficacy correlation corresponding to a certain combination of acoustic stimulation parameters.

[0183] In some embodiments of this application, the system may employ machine learning models, statistical models, rule-learning hybrid models, or multi-model fusion models to train or fit the relationship between historical EEG features, acoustic stimulation parameters, acoustic stimulation signal types, state variables, and therapeutic feedback, thereby generating therapeutic trend prediction results or therapeutic response probabilities.

[0184] The efficacy trend prediction results or efficacy response probability are used to assist doctors in determining the recommended range of subsequent acoustic stimulation parameters or the direction of parameter search space contraction, and are not directly used as disease diagnosis conclusions or final treatment decisions.

[0185] Example 3: Construction of the parameter-signal-state-therapeutic effect mapping relationship

[0186] In one embodiment of this application, the system establishes a parameter-signal-state-therapeutic effect mapping relationship between acoustic stimulation parameters, acoustic stimulation signal types, EEG state variables, EEG objective therapeutic effect indicators, subjective feedback, physiological indicators, and therapeutic effect indicators based on an individual's historical response dataset.

[0187] The mapping relationship is not limited to the result of a black box model, but is a personalized response relationship that can be viewed, reviewed, or traced by doctors.

[0188] For example, the system can analyze the changes in EEG state variables, objective EEG efficacy indicators, user feedback, and efficacy indicators corresponding to a certain type of acoustic stimulus signal and a certain combination of acoustic stimulus parameters at different time periods.

[0189] For example, the system can analyze the relationship between the direction of change of a certain parameter and the changes in the probability of improvement, stabilization, deterioration, or therapeutic response of state variables.

[0190] Mapping relationships can be represented by rule relationships, statistical relationships, weighted scoring relationships, trend relationships, correlation matrices, response curves, or interpretable model parameters.

[0191] Example 4: Artificial Intelligence-Assisted Modeling and Construction of Individual Response Models

[0192] In some embodiments of this application, the system constructs an individual response model based on an individual's historical response dataset and the parameter-signal-state-therapeutic effect mapping relationship. The individual response model is used to describe the trends in the subject's neural state, therapeutic effect feedback, or risk under different acoustic stimulation parameters and different acoustic stimulation signal types.

[0193] Individual response models can be rule-based models, statistical models, machine learning models, rule-learning hybrid models, multi-model fusion models, interpretable scoring models, or trend prediction models. Machine learning models can be used to fit the nonlinear relationships between historical EEG characteristics, acoustic stimulation parameters, acoustic stimulation signal types, state variables, objective EEG efficacy indicators, and efficacy feedback; rule-based models can be used to provide safety boundaries, prohibition conditions, and conservative parameter rules; rule-learning hybrid models can simultaneously utilize historical data to fit results and manually set rules.

[0194] In some embodiments of this application, the AI-assisted modeling and interpretation module can train, fit, predict trends, score risks, and analyze feature contributions on an individual's historical response dataset, and output interpretable assisted recommendation results. The feature contribution analysis can be used to indicate the degree of influence of EEG features, acoustic stimulation parameters, acoustic stimulation signal types, objective EEG efficacy indicators, or user feedback on the recommendation results.

[0195] In some embodiments of this application, the individual response model can output individualized recommendation results for a certain type of acoustic stimulus parameters (individualized parameter recommendation range, recommended acoustic stimulus signal type, prohibited range, recommendation update direction, response score, risk score, efficacy trend prediction result, or efficacy response probability). This output is only used as an auxiliary analysis result and is not directly used as the final treatment decision or automatic diagnosis conclusion.

[0196] Example 5: Personalized Parameter Recommendation and Parameter Search Space Shrinking

[0197] In some embodiments of this application, the system generates individualized parameter recommendation ranges, acoustic stimulus signal type recommendations, parameter priority suggestions, parameter disallowed ranges, efficacy trend prediction results, efficacy response probabilities, or parameter search space shrinkage results based on individual response models.

[0198] For example, when historical data shows that a subject has more stable state variables, better objective EEG efficacy indicators, better user feedback, and fewer safety events under a specific parameter range or a specific acoustic stimulation signal type, the system can set that parameter range and acoustic stimulation signal type as the preferred recommended range; when historical data shows that certain parameter combinations or signal types are associated with adverse feedback, state deterioration, or safety events, the system can list them as disabled parameter ranges or low-priority parameter ranges.

[0199] During subsequent acoustic stimulation interventions, the system can provide individualized recommendations for physician review or confirmation, or allow the acoustic stimulation prescription template module and closed-loop control module to call upon them to limit the parameter search space. By narrowing the parameter search space, the system can reduce invalid parameter attempts and overexploration, thereby improving the efficiency of long-term individualized regulation.

[0200] Example 6: Doctor-side confirmation, safety state machine constraints, and long-term model updates

[0201] In some embodiments of this application, the recommended parameter range, recommended acoustic stimulus signal type, parameter priority suggestions, prohibited parameter range, efficacy trend prediction results, efficacy response probability, or risk warnings output by the individual response model are displayed on the doctor's end. The doctor can confirm, adjust, lock, or reject the model output results.

[0202] In some embodiments of this application, the output of the individual response model can be called by the acoustic stimulus prescription template module to update the initial range of parameters, acoustic stimulus signal type, parameter update priority, safety threshold set, disabling conditions, or fallback rules in the prescription template.

[0203] In other embodiments of this application, the output of the individual response model can be called by the closed-loop control module to generate subsequent parameter update suggestions or limit the parameter search space.

[0204] In the above embodiments, regardless of the calling method, the model output results must be subject to the safety state machine constraint: when the EEG signal quality is insufficient, the state variable fluctuates abnormally, the user feedback is poor, the parameter exceeds the limit, the safety risk increases, or the system operation is abnormal, the system ignores or terminates the model output results and switches to the conservative parameter, pause, stop or rollback mechanism under the safety state machine constraint.

[0205] During subsequent acoustic stimulation interventions, the system continues to record new EEG characteristics, acoustic stimulation parameters, acoustic stimulation signal types, state variables, physiological indicators, subjective feedback, therapeutic efficacy indicators, and objective EEG therapeutic efficacy indicators, and adds them to the individual's historical response dataset.

[0206] When new data meets preset quality and quantity conditions, the system can update the individual response model. During model updates, the system records the input data version, model version, changes in EEG objective efficacy indicators, feature contribution results, changes in recommendation results, physician confirmation records, and safety status switching records. If the model update results in significant changes to recommendation results, increased risk, or lack of physician confirmation, the system can retain the original model version or revert to the previous stable model version.

[0207] The beneficial effects of the present invention, a method and system for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback, are as follows:

[0208] 1. This application can integrate EEG characteristics, sound stimulation parameters, sound stimulation signal types, state variables, physiological indicators, subjective feedback, efficacy indicators and objective EEG efficacy indicators during multiple sound stimulation interventions into an individual historical response dataset, thereby providing a data foundation for long-term individualized optimization;

[0209] 2. This application constructs a parameter-signal-state-therapeutic effect mapping relationship, which can express the individualized association between different acoustic stimulation parameters, different acoustic stimulation signal types, changes in the subject's neural state, objective EEG therapeutic effects indicators, and therapeutic effects feedback;

[0210] 3. This application uses objective EEG efficacy indicators to make efficacy evaluation not limited to subjective feedback or scale evaluation, but can combine changes in EEG characteristic values, changes in EEG state variables, changes in EEG rhythm stability and changes in brain network connectivity characteristics for objective analysis;

[0211] 4. This application uses machine learning models, rule-learning hybrid models, or multi-model fusion models for training, fitting, trend prediction, risk scoring, and feature contribution analysis, which can generate interpretable personalized parameter recommendation ranges, efficacy trend prediction results, or efficacy response probabilities.

[0212] 5. This application uses the output of the artificial intelligence model as an interpretable auxiliary recommendation result, and requires that it be confirmed by the doctor, constrained by the prescription template or the safety state machine before it can be applied, so as to avoid the black box model directly controlling the treatment process;

[0213] 6. In cases of abnormal states, deterioration of state variables, poor user feedback, parameter out-of-bounds errors, or increased security risks, this application can ignore model recommendations and switch to safe state machine constraints to improve system security.

[0214] 7. This application improves the traceability, auditability, and quality management capabilities of long-term acoustic stimulation neuromodulation software by recording model version, data version, feature contribution results, doctor-end confirmation records, and security status switching records.

[0215] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A personalized optimization method for acoustic stimulation neural modulation based on historical EEG and therapeutic feedback, characterized in that, The method establishes a parameter-signal-state-efficacy mapping relationship based on historical EEG data, sound stimulation parameters, sound stimulation signal types, state variables, subjective feedback, efficacy indicators, and objective EEG efficacy indicators obtained during multiple sound stimulation interventions. This relationship is then used to generate personalized recommendation results, and includes the following steps: S1: Obtain historical intervention data, which includes one or more of the following: EEG characteristics, acoustic stimulation parameters, acoustic stimulation signal type, EEG state variables, physiological indicators, subjective feedback, therapeutic efficacy indicators, objective EEG therapeutic efficacy indicators, usage duration, and safety events. S2: Preprocess the historical intervention data to form an individual historical response dataset. The preprocessing includes time stamping, quality screening, anomaly marking, missing data handling, and structured organization. S3: Based on the individual historical response dataset, establish a parameter-signal-state-efficacy mapping relationship between acoustic stimulation parameters, acoustic stimulation signal types, EEG state variables, EEG objective efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators; S4: Based on the parameter-signal-state-therapeutic effect mapping relationship, construct an individual response model. The individual response model is used to describe the trend of changes in the subject's neurological state, therapeutic effect feedback, or risk under different acoustic stimulation parameters and different acoustic stimulation signal types. S5: Generate individualized recommendation results based on the individual response model. The individualized recommendation results include individualized parameter recommendation range, acoustic stimulus signal type recommendation, parameter priority suggestion, parameter disabled range, parameter search space shrinkage result, efficacy trend prediction result or efficacy response probability. S6: Output the individualized recommendation results to the doctor's terminal and the sound stimulation prescription template module or closed-loop control module; S7: After confirmation by the doctor and under the constraints of the prescription template and / or the safety state machine, the individualized recommendation result is used for subsequent acoustic stimulus parameter initialization, acoustic stimulus signal type selection, parameter update suggestion generation, or parameter search space limitation. S8: When insufficient EEG signal quality, abnormal fluctuations in state variables, poor user feedback, parameter out-of-bounds, increased safety risks, or abnormal system operation are detected, ignore the current individualized recommendation results and switch to conservative parameters, pause, stop, or rollback mechanisms under the constraints of the safety state machine. S9: Record new historical intervention data during subsequent acoustic stimulation interventions and use it to update the individual's historical response dataset; S10: Record the construction process of the individual response model, input data version, model version, changes in EEG objective efficacy indicators, feature contribution results, parameter recommendation results, doctor confirmation records, safety status switching records, and model update records.

2. The method for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback as described in claim 1, characterized in that, The EEG characteristics include one or more of the following: power in different frequency bands, relative power, power stability, inter-band ratio, synchronicity, coherence, cross-frequency coupling, brain network connectivity characteristics, rhythm stability, and event-related change characteristics.

3. The method for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback as described in claim 1, characterized in that, The sound stimulation parameters include one or more of the following: sound stimulation type, frequency band range, one or more center frequencies, modulation method, modulation depth, modulation rhythm, sound intensity range, sound stimulation duration, sound stimulation energy, left and right ear output relationship, stimulation interval, and stimulation combination sequence; the sound stimulation signal type includes one or more of the following: pure tone, narrowband noise, broadband noise, modulated sound, rhythmic sound, natural sound, composite sound, musical sound, and pulse stimulation with a set waveform, wherein the modulated sound includes amplitude modulation sound, frequency modulation sound, or a mixture of amplitude modulation and frequency modulation sound.

4. The method for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback as described in claim 1, characterized in that, The physiological indicators include one or more of the following: heart rate, heart rate variability, respiratory-related indicators, sleep-related indicators, activity level, skin conductance, blood oxygenation-related indicators, and body movement indicators; the subjective feedback includes user feedback on auditory stimulation tolerance, discomfort, symptom changes, sleep changes, attention status, fatigue level, emotional state, or user experience; the efficacy indicators include one or more of the following: changes in scale scores, degree of improvement in target state, degree of subjective improvement by the user, changes in sleep quality, changes in symptom burden, changes in compliance, physician follow-up evaluation, and objective efficacy indicators based on historical EEG detection and analysis; the objective efficacy indicators include one or more of the following: changes in EEG characteristic values, changes in EEG state variables, changes in EEG rhythm stability, changes in brain network connectivity characteristics, parameter-EEG-efficacy correlation strength, efficacy trend prediction results, and efficacy response probability.

5. The method for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback as described in claim 1, characterized in that, The parameter-signal-state-treatment mapping relationship is represented by one or more of the following: rule-based relationship, statistical relationship, weighted scoring relationship, trend relationship, correlation matrix, response curve, or interpretable model parameters.

6. The method for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback as described in claim 1, characterized in that, The individual response model is one or more of the following: rule-based model, statistical model, machine learning model, rule-learning hybrid model, and multi-model fusion model; wherein, the machine learning model, rule-learning hybrid model, or multi-model fusion model is used to train, fit, predict trends, score risks, or analyze features of historical EEG characteristics, acoustic stimulation parameters, acoustic stimulation signal types, state variables, objective EEG efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators.

7. The method for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback as described in claim 1, characterized in that, The individualized recommendation results include one or more of the following: recommended parameter range, recommended parameter level, recommended parameter combination, recommended acoustic stimulus signal type, prohibited parameter range, parameter update priority, efficacy trend prediction result, efficacy response probability, and parameter search space boundary.

8. The method for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback according to claim 1, characterized in that, The individual response model does not directly output disease diagnosis conclusions, nor is it directly used as the final treatment decision; the output results of the individual response model are used to assist doctors in confirmation, prescription template adjustment, or parameter updates under safety state machine constraints.

9. The method for individualized optimization of acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback according to claim 1, characterized in that, The security state machine constraints include one or more of the following: signal quality gating, security threshold verification, disable condition judgment, observation period constraints, rollback rules, pause rules, and stop rules.

10. A personalized optimization system for acoustic stimulation neuromodulation based on historical EEG and therapeutic feedback, characterized in that, include: The historical data acquisition module is used to collect or record historical intervention data; The data processing module is used to perform time stamping, quality screening, anomaly marking, missing data handling, and structured processing on historical intervention data to form individual historical response datasets; The mapping relationship construction module is used to establish parameter-signal-state-efficacy mapping relationships between acoustic stimulation parameters, acoustic stimulation signal types, EEG state variables, EEG objective efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators. The individual response model building module is used to build an individual response model based on the individual's historical response dataset and the parameter-signal-state-treatment mapping relationship; The AI-assisted modeling and interpretation module is used to train, fit, predict trends, score risks, and analyze feature contributions of EEG features, acoustic stimulation parameters, acoustic stimulation signal types, state variables, objective EEG efficacy indicators, subjective feedback, physiological indicators, and efficacy indicators in an individual's historical response dataset based on machine learning models, rule-learning hybrid models, or multi-model fusion models, and output interpretable auxiliary recommendation results. The parameter recommendation module is used to generate personalized recommendation results based on the individual response model; The doctor-side confirmation module is used to display or review the individualized recommendation results and record doctor-side confirmation, adjustment, locking, or rejection information; The safety constraint module is used to ignore the current individualized recommendation results and switch to conservative parameters, pause, stop, or rollback mechanisms when an abnormal state or safety risk is detected. The model update and log module is used to record the model building process, input data version, model version, changes in EEG objective efficacy indicators, feature contribution results, parameter recommendation results, doctor confirmation records, safety status switching records, and model update records.

Images